首页|基于Grid-Search的Dropout-LSTM模型在新冠肺炎预测中的应用

基于Grid-Search的Dropout-LSTM模型在新冠肺炎预测中的应用

扫码查看
为了对新冠肺炎(COVID-19)实现更准确地分析与预测,建立了基于LSTM(long short-term memory)的新冠肺炎预测模型,使用网格搜索法对最关键的3个超参数进行优化.同时,为了提高预测准确性,解决过拟合现象,引入Dropout正则化对网络进行了优化.实验证明:具有多层隐藏层的LSTM模型比传统LSTM模型有更高的预测准确性;当Dropout比率为0.22时,能够有效地解决模型预测时出现的过拟合问题;相比较RNN(recurrent neural network)模型和SEIR(susceptible ex-posed infected recovered)模型,所建立的基于Grid-Search的Dropout-LSTM模型的均方根误差、平均绝对误差和平均绝对百分比误差均最小.因此,建立的基于Grid-Search的Dropout-LSTM模型有更优的预测能力.
Application of Grid-Search Based Dropout-LSTM model in the Prediction of COVID-19
To achieve more accurate analysis and prediction of the epidemic situation of COVID-19 2019(COVID-19),an LSTM(long-and short-term memory)based prediction model of COVID-19 is established,and the grid search method is used to opti-mize the three most critical super parameters.At the same time,in order to improve the prediction accuracy and solve the over-fitting phenomenon,the Dropout regularization is introduced to optimize the network.Testing shows that the LSTM model with multiple hidden layers has higher prediction accuracy than the traditional LSTM model.When the dropout ratio is 0.22,it can effectively solve the problem of over-fitting in model prediction.Compared with RNN(Recurrent Neural Network)model and SEIR(susceptible exposed infected recovered)model,the Dropout-LSTM model established in this experiment based on grid-search has the smallest root mean square error,average absolute error and average absolute percentage error.Therefore,the Dropout-LSTM model based on grid-search established in this experiment has better prediction ability.

COVID-19LSTMneural networkforecastRNN

齐悦、谢泰、沙琨

展开 >

海军军医大学,海军卫生信息中心,上海 200082

新冠肺炎 LSTM 神经网络 预测 RNN

2024

微型电脑应用
上海市微型电脑应用学会

微型电脑应用

CSTPCD
影响因子:0.359
ISSN:1007-757X
年,卷(期):2024.40(2)
  • 13